Last year, neuroscientist Jen Ware accepted a newly established position as director of experimental design at the CHDI Foundation in New York City, a non-profit organization that seeks treatments for Huntington's disease. She helps CHDI-funded scientists to plan and conduct robust research.

How did CHDI come to create this role?

The low reproducibility of findings in research misdirects effort and wastes a lot of money. My job stems from CHDI's concern for scientific rigour and quality. That concern came to a head last year when the president of the foundation was having a chat with Marcus Munafò, my postdoc adviser at the University of Bristol, UK. He had done a meta-analysis of neuroscience studies that showed how studies with inadequate sample sizes can lead to spurious results — false positives as well as false negatives (K. S. Button et al. Nature Rev. Neurosci. 14, 365–376; 2013).

What do you do in your job?

One duty is to coordinate review of study protocols by an independent committee and to get feedback to scientists before studies are run. It's feedback on whether the methods and the sample size are adequate for the research question. Another duty is statistics and methods training for postdocs in the labs we work with. We hope to set up online courses that postdocs can complete and possibly receive certification for.

Do researchers worry that this will make more work for them?

I'm introducing policies so that protocols will be reviewed. It might be a little more effort at the start of a project, but when it concludes, the research will be easier to assess. I'm trying to make it easier going forward.

Tell me about your background.

It's been a winding career path. I started with a bachelor's degree in psychology at Cardiff University, UK, and then I decided to work as a substance-abuse counsellor. I was on a team that sets up services and reparation for youth, and I had unique insights into drug dependency from working on the front line. So I decided to do a PhD in neuroscience.

What led you to experimental design?

I ran a genome-wide analysis of cotinine levels in smokers for my PhD. Cotinine is the primary metabolite of nicotine, and it's a more precise measure of smoking heaviness than metrics such as self-reporting. We had a sample size of only 4,500, but our results were comparable with studies that have used samples 4 or 5 times that size but with less-precise measures. Later, I spent time at Stanford University in California working on biases in MRI studies with John Ioannidis, who researches scientific rigour. I also heard stories from postdocs who had been told that it was their job to find significant relationships in a data set — to essentially go P-value fishing. These factors contributed to my interest in meta-research — research on research — and scientific rigour.

What did you learn about statistical analysis?

I had to learn that P values by themselves aren't particularly meaningful. You can get a result with a very low P value, but it could be related to an effect that has no clinical or biological relevance. We need to consider whether studies include enough observations to detect a meaningful difference between two groups. Non-significant results of an adequately powered study can be more meaningful than significant results from an underpowered study.

Were you concerned about being the first person in a new position?

I had no trepidation in accepting the offer, but I would be lying if I said that I wasn't a little daunted. It was a move from academia to a non-profit foundation, and a move from the United Kingdom to the United States. It was a lot of change all at once. But I'm already feeling that this is the right move for me.

This interview has been edited for length and clarity.